PR #8634 Optimize quantized LSTM

Optimize quantized LSTM #8634

zhanghuanrong merged 11 commits into master from zhalei/qlstm3

optimize some lstm gate computation. Remove no need string constructi…

71da820c

change gcc optimization flags for computation bound logics in rnn_hel…

30f7e8a3

better qgemm for M=1

5dc2f787

Some improve on avx512

a0cf1131

zhanghuanrong requested a review from

yufenglee 4 years ago

zhanghuanrong requested a review from

tracysh 4 years ago

zhanghuanrong requested a review 4 years ago

add condition to limit GCC related marcros

aa21fe8f

Correct QGemm assembly for M=1 AVX2 optimization to pass mlas_test.

04c3d924

Fix rnn_helper build issue for wasm.

ab7ed387

better asm code here according to feedbacks.

c8508dec

Remove customized vectorize and unroll option for GCC.

c985ad68

Better restrict semantic for merge_lstm_gates_to_memory() by adding i…

5fcb50ad

yufenglee changed the title ~~Zhalei/qlstm3~~ Optimize quantized LSTM 4 years ago

yufenglee dismissed these changes on 2021-08-10

Force CI restart as it stucked by the onnxruntime-python-checks-ci-pi…

de1b7ba0

zhanghuanrong dismissed their stale review via de1b7ba0 4 years ago

yufenglee approved these changes on 2021-08-12

zhanghuanrong merged 76dfe810 into master 4 years ago

zhanghuanrong deleted the zhalei/qlstm3 branch 4 years ago

Reviewers

yufenglee

tracysh

Assignees

No one assigned

Labels

None yet

Milestone

No milestone